Multi-Functional Beamforming Design for Integrated Sensing, Communication, and Computation

TL;DR

This paper proposes a novel multi-functional system integrating sensing, communication, and computation, optimizing resource allocation to improve radar target estimation performance while satisfying user and computational constraints.

Contribution

It introduces new algorithms based on SDR for joint optimization in an integrated sensing, communication, and computation system, enhancing target estimation accuracy.

Findings

Proposed algorithms outperform benchmarks in target estimation accuracy.

Efficient resource allocation balances sensing, communication, and computation needs.

Optimal rank-one solutions are obtainable even with multiple functionalities.

Abstract

Integrated sensing and communication (ISAC) systems may face a heavy computation burden since the sensory data needs to be further processed. This paper studies a novel system that integrates sensing, communication, and computation, aiming to provide services for different objectives efficiently. This system consists of a multi-antenna multi-functional base station (BS), an edge server, a target, and multiple singleantenna communication users. The BS needs to allocate the available resources to efficiently provide sensing, communication, and computation services. Due to the heavy service burden and limited power budget, the BS can partially offload the tasks to the nearby edge server instead of computing them locally. We consider the estimation of the target response matrix, a general problem in radar sensing, and utilize Cramer-Rao bound (CRB) as the corresponding performance metric. To tackle the non-convex optimization problem, we propose both semidefinite relaxation (SDR)-based alternating optimization and SDR-based successive convex approximation (SCA) algorithms to minimize the CRB of radar sensing while meeting the requirement of communication users and the need for task computing. Furthermore, we demonstrate that the optimal rankone solutions of both the alternating and SCA algorithms can be directly obtained via the solver or further constructed even when dealing with multiple functionalities. Simulation results show that the proposed algorithms can provide higher target estimation performance than state-of-the-art benchmarks while satisfying the communication and computation constraints.